Skip to content
This repository has been archived by the owner on Sep 18, 2024. It is now read-only.

Add recently-idle environment scheduler in reuse mode #3375

Merged
merged 150 commits into from
Feb 24, 2021

Conversation

SparkSnail
Copy link
Contributor

No description provided.

SparkSnail and others added 30 commits May 29, 2020 17:02
1. rename storage file name
2. add more log on status changes
3. change isEnd to isAlive for better naming
add internal prefix for internal storage methods for clear usage.
fix pylint errors
minor fixes
rename methods of storageService
move trial to a seperated file
fix some bugs.
fix openPAI breaking changes
fix minor bugs
 to router training service for better understanding.
trialService is used to support different submission types like AML.
TrialDispatcher is easier to understand it's purpose.
@J-shang
Copy link
Contributor

J-shang commented Feb 18, 2021

please fix the lint

@J-shang J-shang mentioned this pull request Feb 19, 2021
94 tasks
@@ -30,7 +30,7 @@ export class GpuScheduler {

// private readonly machineExecutorMap: Set<TrialDetail>;
private readonly log: Logger = getLogger();
private readonly policyName: SCHEDULE_POLICY_NAME = 'round-robin';
private readonly policyName: SCHEDULE_POLICY_NAME = 'recently-idle';
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is it possible that only apply this policy for aml training service?

if (selectedEnvironment === undefined) {
return this.roundRobinSelect(qualifiedEnvironments, allEnvironments);
}
selectedEnvironment.latestTrialReleasedTime = -1;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

BTW, one environment only runs one trial at a time?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if it is, then policy looks good. if not, one environment can run two trials concurrently, then this policy becomes round robin

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It could run multiple trial at one environment, why it will become round-robin?

@SparkSnail SparkSnail closed this Feb 24, 2021
@SparkSnail SparkSnail reopened this Feb 24, 2021
@SparkSnail
Copy link
Contributor Author

please fix the lint

fixed.

@J-shang J-shang merged commit cd05da6 into microsoft:master Feb 24, 2021
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants